[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$fpM7pc-IfW6Xj1Qll7dbjynFK9EK8SwI_-1CvirOurDE":3,"$fm6Ky0r9HHO8PTXKBgXJeU4fCLm4z6RKRMXFRXG3f560":26},{"success":4,"data":5},true,{"id":6,"slug":7,"title":8,"excerpt":9,"content":10,"category":11,"tags":12,"author":20,"cover_image_url":21,"reading_time_minutes":22,"is_published":4,"published_at":23,"created_at":24,"updated_at":24,"author_avatar":21,"is_featured":25,"meta_title":21,"meta_description":21,"meta_keywords":21},"4d06101f-bc96-4090-a72b-9798e3018e04","colorado-ai-hiring-law-comment-deadline","Colorado Rewrote Its AI Hiring Law — Public Comments Close in Two Days","Colorado replaced its 2024 AI Act with SB 26-189, a narrower notice-disclosure-correction-review framework covering hiring decisions. Draft rules are open for public comment until 4 September 2026, with a hearing 26 October, ahead of the 1 January 2027 effective date.","\u003Cp>Colorado has spent two years trying to finish a single piece of AI legislation. The state's original AI Act — Senate Bill 24-205, signed in 2024 — was supposed to take effect in February 2026, then June 2026, after two separate legislative delays. In early 2026, a lawsuit from xAI temporarily suspended enforcement altogether. On 14 May 2026, Governor Jared Polis signed the resolution to that saga: SB 26-189, which repeals SB 24-205 outright and re-enacts a narrower law in its place. It takes effect \u003Cstrong>1 January 2027\u003C\u002Fstrong>.\u003C\u002Fp>\n\n\u003Cp>That would already be worth a hiring team's attention. What makes it timely this week is what has happened since: the Colorado Attorney General's office published draft implementing rules on 11 August 2026, and the public comment window on those rules closes \u003Cstrong>4 September 2026\u003C\u002Fstrong> — two days from this piece going up. A second, revised draft is due by 23 September, and a public hearing is scheduled for 26 October 2026 in Denver. If your hiring stack uses AI to screen, rank, or evaluate candidates and you employ people in Colorado, this is the moment the actual operating rules are still being written, not just the statute.\u003C\u002Fp>\n\n\u003Ch2>What changed, and why it matters that it's less\u003C\u002Fh2>\n\n\u003Cp>SB 24-205 was, by design, one of the most demanding AI laws in the United States: a duty of care to prevent algorithmic discrimination, mandatory risk-management programs, pre-deployment impact assessments, and reporting obligations to the Attorney General. SB 26-189 removes all of that. In its place is a narrower, disclosure-and-correction framework built around four deployer duties, consistently described the same way across the legal analyses that have covered the bill — from \u003Ca href=\"https:\u002F\u002Fwww.finnegan.com\u002Fen\u002Finsights\u002Farticles\u002Fcolorado-replaces-landmark-ai-act-an-overview-of-the-new-sb-26-189-framework.html\">Finnegan\u003C\u002Fa>, \u003Ca href=\"https:\u002F\u002Fogletree.com\u002Finsights-resources\u002Fblog-posts\u002Fcolorados-new-ai-act-targets-automated-decision-making-for-consequential-decisions\u002F\">Ogletree Deakins\u003C\u002Fa>, and \u003Ca href=\"https:\u002F\u002Fwww.buchalter.com\u002Finsights\u002Fcolorado-rewrites-its-ai-law-what-employers-must-know-about-sb-26-189\u002F\">Buchalter\u003C\u002Fa>, among others:\u003C\u002Fp>\n\n\u003Col>\n\u003Cli>\u003Cstrong>Point-of-interaction notice.\u003C\u002Fstrong> A deployer must give a consumer clear, conspicuous notice before using a covered automated decision-making technology (ADMT) to materially influence a consequential decision about them.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Post-adverse-outcome disclosure.\u003C\u002Fstrong> When an ADMT contributes to an adverse outcome, the deployer has 30 days to give the affected person a plain-language explanation of the decision and the technology's role in it.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Correction rights.\u003C\u002Fstrong> The individual can request correction of factually inaccurate personal data that was used in the decision.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Meaningful human review.\u003C\u002Fstrong> The individual can request human review and reconsideration of the adverse outcome, to the extent commercially and technically reasonable.\u003C\u002Fli>\n\u003C\u002Fol>\n\n\u003Cp>There's also a three-year recordkeeping duty behind all four. The law folds in a fifth Colorado bill worth naming alongside it — HB 26-1263, the Chatbot Safety Act — which the Attorney General's office is implementing through the same rulemaking docket, since a growing share of \"automated decision-making\" in hiring now arrives through conversational AI rather than a static scoring model.\u003C\u002Fp>\n\n\u003Cp>Crucially, \"consumer\" under SB 26-189 is defined broadly enough to reach the workplace directly: it includes employees and job applicants who are Colorado residents, plus anyone whose access or eligibility in Colorado is evaluated in a \"consequential decision.\" Employment sits explicitly alongside education, lending, insurance, health care, housing, and essential government services as one of the law's seven covered domains — and the statute's own language for the employment leg is broad: hiring, screening, promotion, compensation, discipline, and termination decisions all qualify when an ADMT \"materially influences\" them.\u003C\u002Fp>\n\n\u003Ch2>The rulemaking is where the real detail lives right now\u003C\u002Fh2>\n\n\u003Cp>A statute this compressed leaves a lot to regulation, and the Attorney General's office has used that room. The draft rules released 11 August, according to coverage from \u003Ca href=\"https:\u002F\u002Fwww.bytebacklaw.com\u002F2026\u002F08\u002Fcolorado-attorney-general-releases-proposed-rules-for-the-new-admt-act\u002F\">Byte Back Law\u003C\u002Fa> and \u003Ca href=\"https:\u002F\u002Fwww.seyfarth.com\u002Fnews-insights\u002Fcolorado-releases-proposed-rules-for-its-ai-and-chatbot-safety-laws-these-create-more-operational-work-than-the-statutes-suggestlegalupdate.html\">Seyfarth Shaw\u003C\u002Fa>, go well beyond restating the statute: they attempt to define when an ADMT \"materially influences\" a decision, what a compliant post-adverse-outcome disclosure has to contain in practice, and what counts as \"meaningful\" human review rather than a rubber stamp. Both write-ups flag the same theme from opposite angles — the rules create considerably more day-to-day operational work than the four-duty summary of the statute suggests, particularly for anyone who has to build the actual notice and disclosure copy, log it, and route correction and review requests to a real person within a defined window.\u003C\u002Fp>\n\n\u003Cp>That draft is not final. Comments submitted by 4 September go into a revised version circulated by 23 September, ahead of the 26 October hearing — which the Attorney General's office is running jointly for the ADMT rules and the Chatbot Safety Act rules (see \u003Ca href=\"https:\u002F\u002Fcoag.gov\u002Fai\u002F\">coag.gov\u002Fai\u003C\u002Fa>). For any organization with Colorado hiring volume, this two-week window is the last point where the content of \"meaningful human review\" is still being argued, not just implemented.\u003C\u002Fp>\n\n\u003Ch2>Where this sits in the wider patchwork\u003C\u002Fh2>\n\n\u003Cp>Colorado is not writing on a blank page nationally. New York City's Local Law 144 bias-audit requirement has applied to hiring tools since 2023. California's Civil Rights Council automated-decision-system regulations took effect in October 2025. Illinois has its own AI hiring notification law. What's distinctive about Colorado's approach — even in its scaled-back form — is that it's the first state framework to pair notice and disclosure duties with an explicit, individually enforceable correction right and a human review right, rather than leaving those as best practice. A candidate doesn't just get told an AI was involved; under the statute, they get a channel to say \"that's wrong\" and a channel to ask a person to look again.\u003C\u002Fp>\n\n\u003Cp>That is a meaningfully different compliance shape than an audit report filed once a year. It's closer to a standing operational obligation: every deployer needs a working answer, on the day a rejection goes out, to \"who reviews this if the candidate asks,\" not just a paper trail proving the tool was tested for bias last spring.\u003C\u002Fp>\n\n\u003Ch2>Our reading, and where we're honest about the gap\u003C\u002Fh2>\n\n\u003Cp>Two of these four duties describe infrastructure we've already had to build for a different reason — the EU AI Act's Article 50 transparency duty, which we covered \u003Ca href=\"https:\u002F\u002Fwww.nicehire.ai\u002Fblog\u002Feu-ai-act-august-2026-what-applies-hiring\">here\u003C\u002Fa> in early August. Every greeting path in NiceHire's AI screening flow — Tavus and Retell voice, every generated and preset template, in English, Japanese, Traditional and Simplified Chinese — is routed through a single non-suppressible disclosure function, \u003Ccode>ensureAIDisclosure()\u003C\u002Fcode>, so a custom greeting can change the wording but cannot remove the AI self-identification. That's the same shape as Colorado's point-of-interaction notice duty, even though we built it for a different regulator. Completed screening interviews also record a transcript against the application, which is the raw material a post-adverse-outcome disclosure would draw from — with two honest limits: the transcript is whatever the interview provider actually returns, so a session that ends without one is stored without one, and re-running a screening clears the prior attempt's transcript rather than keeping both.\u003C\u002Fp>\n\n\u003Cp>What we do not have, and want to say plainly rather than imply: a built correction workflow letting a candidate flag inaccurate data used in a screening decision, or a defined \"meaningful human review\" path with a service-level response window. Those are the third and fourth duties, and they are also the two duties the legal commentary above singles out as the harder operational lift — because they require a real person on the other end, on a schedule, not just a log entry. Colorado's rulemaking closing on 4 September is, among other things, the process that will define how demanding that lift actually needs to be. We're watching it for the same reason any employer using AI in hiring should be: the four-duty summary is stable, but what satisfies duty three and four in practice is still being written this month.\u003C\u002Fp>\n\n\u003Cp>For a hiring team with Colorado headcount, our practical list going into Q4:\u003C\u002Fp>\n\n\u003Col>\n\u003Cli>\u003Cstrong>Inventory\u003C\u002Fstrong> every point in your hiring flow where a tool scores, ranks, or filters candidates — including features embedded in vendor products you didn't build yourself. The \"materially influences\" test is going to be argued at the tool level, not the platform level.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Ask each vendor, in writing, who owns duties three and four.\u003C\u002Fstrong> A tool that discloses AI use but has no path for correction or human review only gets you halfway to compliant.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Watch the 23 September revised draft\u003C\u002Fstrong>, not just the 11 August one — the Attorney General's office has said comments will shape it, and the gap between draft and final is exactly where \"meaningful\" gets defined.\u003C\u002Fli>\n\u003Cli>\u003Cstrong>Build your January 2027 rollout around the hearing date\u003C\u002Fstrong>, not the effective date. 26 October leaves roughly nine weeks to operationalize whatever the final rule says — tighter than it looks from the 1 January deadline alone.\u003C\u002Fli>\n\u003C\u002Fol>\n\n\u003Cp>Sources: \u003Ca href=\"https:\u002F\u002Fwww.finnegan.com\u002Fen\u002Finsights\u002Farticles\u002Fcolorado-replaces-landmark-ai-act-an-overview-of-the-new-sb-26-189-framework.html\">Finnegan\u003C\u002Fa>, \u003Ca href=\"https:\u002F\u002Fogletree.com\u002Finsights-resources\u002Fblog-posts\u002Fcolorados-new-ai-act-targets-automated-decision-making-for-consequential-decisions\u002F\">Ogletree Deakins\u003C\u002Fa>, \u003Ca href=\"https:\u002F\u002Fwww.buchalter.com\u002Finsights\u002Fcolorado-rewrites-its-ai-law-what-employers-must-know-about-sb-26-189\u002F\">Buchalter\u003C\u002Fa>, \u003Ca href=\"https:\u002F\u002Fwww.bytebacklaw.com\u002F2026\u002F08\u002Fcolorado-attorney-general-releases-proposed-rules-for-the-new-admt-act\u002F\">Byte Back Law\u003C\u002Fa>, \u003Ca href=\"https:\u002F\u002Fwww.seyfarth.com\u002Fnews-insights\u002Fcolorado-releases-proposed-rules-for-its-ai-and-chatbot-safety-laws-these-create-more-operational-work-than-the-statutes-suggestlegalupdate.html\">Seyfarth Shaw\u003C\u002Fa>, and the Colorado Attorney General's rulemaking docket at \u003Ca href=\"https:\u002F\u002Fcoag.gov\u002Fai\u002F\">coag.gov\u002Fai\u003C\u002Fa>.\u003C\u002Fp>\n\n\u003Cp>\u003Cem>Byline: NiceHire Team\u003C\u002Fem>\u003C\u002Fp>","Compliance",[13,14,15,16,17,18,19],"AI hiring","Colorado","AI regulation","compliance","ADMT","US employment law","hiring technology","NiceHire Team",null,7,"2026-09-02T00:17:55.903+00:00","2026-09-02T00:17:56.160368+00:00",false,{"success":4,"data":27},{"posts":28,"count":74,"hasMore":4},[29,31,47,61],{"id":6,"slug":7,"title":8,"excerpt":9,"category":11,"tags":30,"author":20,"cover_image_url":21,"reading_time_minutes":22,"published_at":23},[13,14,15,16,17,18,19],{"id":32,"slug":33,"title":34,"excerpt":35,"category":11,"tags":36,"author":20,"cover_image_url":21,"reading_time_minutes":45,"published_at":46},"78d71538-b8f8-47e6-b802-ac95dfc9f3a5","ai-hiring-vendor-liability-mobley-eightfold","When the Screening Tool Becomes the Defendant: What Mobley v. Workday and the Eightfold FCRA Suit Mean for AI Hiring Vendors","Two US cases test whether an AI hiring vendor — not just the employer that bought its software — can be sued directly. Mobley v. Workday established a screening vendor can be liable as an employer's “agent”; Kistler v. Eightfold AI argues a scoring platform is an unregistered credit-reporting agency under the FCRA. Neither is decided, but both target the hidden score. We cover where the cases stand, and where NiceHire's own scoring and rejection defaults sit against the same questions.",[13,37,38,39,40,41,42,43,44],"employment law","FCRA","EEOC","Workday","Eightfold AI","hiring compliance","AI vendor liability","HR tech",9,"2026-08-31T00:21:11.721+00:00",{"id":48,"slug":49,"title":50,"excerpt":51,"category":11,"tags":52,"author":20,"cover_image_url":21,"reading_time_minutes":59,"published_at":60},"5e38e0aa-f935-45bd-8e03-e0391b5d085d","us-ai-hiring-law-patchwork-colorado-retreat","Colorado Gutted Its AI Hiring Law. The Patchwork Around It Got Stricter Anyway.","Colorado just stripped its flagship AI employment law down to a disclosure-only regime and delayed it to 2027. California, Illinois, and New York City obligations kept advancing on schedule anyway. Here is what is actually enforceable for AI hiring tools right now, and why disclosure and recordkeeping are the two requirements every rewrite keeps landing on.",[13,53,37,54,55,56,57,15,58,19],"Colorado AI Act","HR compliance","California employment law","Illinois HB 3773","NYC Local Law 144","automated decision systems",8,"2026-08-30T00:15:58.51+00:00",{"id":62,"slug":63,"title":64,"excerpt":65,"category":11,"tags":66,"author":20,"cover_image_url":21,"reading_time_minutes":59,"published_at":73},"4fe41550-5417-498a-8e69-ba0005702bc3","pay-transparency-patchwork-2026","The Pay Transparency Map Just Got More Crowded: What Maine, Virginia, and Delaware Mean for Hiring in 2026","Virginia's and Maine's salary-range laws took effect in the last two months; Delaware's arrives in 2027. Massachusetts and New Jersey show what real enforcement looks like once the grace period ends. A look at the growing state patchwork, the strictest-law trap remote hiring creates, and a practical checklist for any team posting jobs across state lines.",[67,16,68,44,69,70,71,72],"pay transparency","hiring law","remote hiring","talent acquisition","salary transparency","2026 regulations","2026-08-26T00:27:37.929+00:00",14]